Skip to content

anaconda.md

Latest commit

 

History

History
45 lines (40 loc) · 2.26 KB

anaconda.md

File metadata and controls

45 lines (40 loc) · 2.26 KB

Installing Anaconda2

Xilinx recommends using Anaconda2 to operate in a virtual environment:

  1. Download Anaconda2
    wget https://repo.anaconda.com/archive/Anaconda2-5.1.0-Linux-x86_64.sh
  2. Run the installer (Installer requires bzip, please install it if you don't have it)
    bash ./Anaconda2-5.1.0-Linux-x86_64.sh
  3. Ensure that your .bashrc is preparing Anaconda, by including these lines
    ~/.bashrc: export PATH=/home/<user>/anaconda2/bin:$PATH
    ~/.bashrc: . /home/<user>/anaconda2/etc/profile.d/conda.sh
  4. After updating the bashrc source it to load the new anaconda path
    source ~/.bashrc
  5. As a precaution unset PYTHONPATH to avoid conflicts with packages on your rootfs
    unset PYTHONPATH

Create ml-suite Anaconda Virtual Environment

  1. Invoke bash (if you aren't already in bash)
    bash
  2. Create Virtual Environment
    conda create --name ml-suite python=2.7 numpy=1.14.5 x264=20131218 caffe pydot pydot-ng graphviz keras scikit-learn -c conda-forge
  3. Fix symbolic links between pre-compiled Caffe (libcaffe.so), and OpenCV
    • Note: If you installed anaconda in a custom location, you will need to modify the first line of this script
      bash fix_caffe_opencv_symlink.sh
  4. Activate Environment
    conda activate ml-suite
  5. Verify your environment by importing caffe in python
    python -c "import caffe"
  6. Install TensorFlow version 1.8
    pip install tensorflow==1.8
  7. Install Jupyter to the ml-suite env
    pip install jupyter

Notes for running on AWS w/ Anaconda

At the moment, AWS requires root privileges to deploy on FPGAs. Follow the steps below as a work around:

  1. Become root sudo su
  2. Set Environment Variables Required by runtime source <MLSUITE_ROOT>/overlaybins/setup.sh aws
  3. Set User Environment Variables Required to run Anaconda source ~centos/.bashrc
  4. Activate the users Anaconda Virtual Environmentsource activate ml-suite
  5. The environment is setup and ready to run applications and examples.

Notes for running on NIMBIX w/ Anaconda

  1. Set User Environment Variables Required to run Anaconda source ~/.bashrc
  2. Activate the users Anaconda Virtual Environmentsource activate ml-suite
  3. The environment is setup and ready to run applications and examples.